release-clm5.0.30 merging with ctsm1.0.dev093 prior to fates_next_api merge

[glemieux]

Creating this issue as a place to discuss issues with merging release-clm5.0.30 and ctsm1.0.dev093 as noted here:
ESCOMP#1008 (comment)

Deconflicting branch is https://github.com/glemieux/ctsm/tree/release-clm5.0.30-ctsm1.0.dev093

[glemieux]

@ekluzek doing a diff on the two branches I notice that bld/configure is not included in the dev093 tag on master. Did this code get renamed or moved elsewhere?

[ekluzek]

It was removed on master and is unneeded.

[glemieux]

@ekluzek the following exists in the release branch version of the code. Should it be included in the merge up to ctsm1.0.dev093? Note I have not corrected the change from waterflux_inst% to water_inst%waterfluxbulk_inst% yet:

ctsm/src/main/lnd2atmMod.F90

Lines 364 to 371 in 8ac9d0b

	! NOTE(wjs, 2018-12-05) Similarly, it's not entirely appropriate to put
	! qflx_runoff_rain_to_snow_conversion_col into qflx_qrgwl_col, since again, this
	! is generated over landunits other than glaciers, wetlands and lakes. I'm
	! putting it in qflx_qrgwl_col partly because glacier landunits are typically
	! the primary source of this flux, and partly just because I'm not sure where
	! else to put it.
	waterflux_inst%qflx_qrgwl_col(c) = waterflux_inst%qflx_qrgwl_col(c) + &
	waterflux_inst%qflx_runoff_rain_to_snow_conversion_col(c)

The commit for this code is: ESCOMP@71b3a0a

[ekluzek]

glacier_region_rain_to_snow_behavior was something that was JUST put on the release branch. So we should remove it in this update of fates_next_api, so that fates_next_api will be like what's on master.

[glemieux]

glacier_region_rain_to_snow_behavior was something that was JUST put on the release branch. So we should remove it in this update of fates_next_api, so that fates_next_api will be like what's on master

Does that mean that everything in the above commit should be reverted: ESCOMP@71b3a0a? Including things like rain_to_snow_runoff?

[ekluzek]

Does that mean that everything in the above commit should be removed: ESCOMP@71b3a0a?

Yes.

release-clm5.0.15 was skipped on master (the above commit you point to). I don't see anything else that's like that (there was a cime update that doesn't apply for master in release-clm5.0.19, but nothing like this other commit)

[glemieux]

This branch passes regression test comparisons using clm_short now. Results are here:

/glade/u/home/glemieux/scratch/ctsm-tests/tests_ctsmdev-release-merge

[glemieux]

Tested this branch last weekend with full aux_clm regression suite:

/glade/u/home/glemieux/scratch/ctsm-tests/tests_ctsmdev-release-merge-fullregression

I'm getting the following DIFF fails:

    FAIL ERP_P36x2_Lm13.f10_f10_musgs.IHistClm50Bgc.cheyenne_gnu.clm-monthly BASELINE ctsm1.0.dev093: DIFF
    FAIL ERS_Ly3_Mmpi-serial.1x1_smallvilleIA.IHistClm50BgcCropQianGs.cheyenne_gnu.clm-cropMonthOutput BASELINE ctsm1.0.dev093: DIFF
    FAIL SMS_Ly1_Mmpi-serial.1x1_brazil.IHistClm50BgcQianGs.cheyenne_gnu.clm-output_bgc_highfreq BASELINE ctsm1.0.dev093: DIFF
    FAIL ERP_D_Ld3.f09_g17.I2000Clm50SpGs.cheyenne_intel.clm-prescribed BASELINE ctsm1.0.dev093: DIFF
    FAIL ERP_D_Ld5.f10_f10_musgs.IHistClm50BgcCrop.cheyenne_intel.clm-allActive BASELINE ctsm1.0.dev093: DIFF
    FAIL ERP_Ly3_P72x2.f10_f10_musgs.IHistClm50BgcCrop.cheyenne_intel.clm-cropMonthOutput BASELINE ctsm1.0.dev093: DIFF
    FAIL ERP_P36x2_Lm13.f10_f10_musgs.IHistClm50Bgc.cheyenne_intel.clm-monthly BASELINE ctsm1.0.dev093: DIFF
    FAIL ERS_Ly3_P72x2.f10_f10_musgs.IHistClm50BgcCropG.cheyenne_intel.clm-cropMonthOutput BASELINE ctsm1.0.dev093: DIFF
    FAIL ERS_Ly5_P144x1.f10_f10_musgs.IHistClm50BgcCrop.cheyenne_intel.clm-cropMonthOutput BASELINE ctsm1.0.dev093: DIFF
    FAIL ERS_Ly5_P72x1.f10_f10_musgs.IHistClm45BgcCrop.cheyenne_intel.clm-cropMonthOutput BASELINE ctsm1.0.dev093: DIFF
    FAIL ERS_Ly6_Mmpi-serial.1x1_smallvilleIA.IHistClm50BgcCropQianGs.cheyenne_intel.clm-cropMonthOutput BASELINE ctsm1.0.dev093: DIFF
    FAIL LCISO_Lm13.f10_f10_musgs.IHistClm50BgcCrop.cheyenne_intel.clm-ciso_monthly BASELINE ctsm1.0.dev093: DIFF
    FAIL SMS_D_Ly6_Mmpi-serial.1x1_smallvilleIA.IHistClm45BgcCropQianGs.cheyenne_intel.clm-cropMonthOutput BASELINE ctsm1.0.dev093: DIFF

Note that all the DIFF failures, except one, are for the compsets IHistClm all of which also failed NLCOMP. All NLCOMP failure are for this compset timeframe as well, although note that not all NLCOMP failures resulted in a DIFF failure (i.e. some PASSED baseline check).

All but one of the IHistClm DIFF failures is a BgcCrop compset. This reminded me that there was an unresolved difference in the namelist_ctsm_default.xml that might be related? In the ctsm1.0.dev093 baseline the following lines have irrigate=.false:

clm4.5:

ctsm/bld/namelist_files/namelist_defaults_ctsm.xml

Lines 532 to 534 in ce3bab4

	<init_interp_attributes sim_year="1850" use_cndv=".false." use_fates=".false." lnd_tuning_mode="clm4_5_cam6.0"
	>hgrid=0.9x1.25 maxpft=79 mask=gx1v7 use_cn=.true. use_nitrif_denitrif=.true. use_vertsoilc=.true. use_crop=.true. irrigate=.true. glc_nec=10
	</init_interp_attributes>

clm5.0:

ctsm/bld/namelist_files/namelist_defaults_ctsm.xml

Lines 554 to 556 in ce3bab4

	<init_interp_attributes sim_year="1850" use_cndv=".false." use_fates=".false." lnd_tuning_mode="clm5_0_cam6.0"
	>hgrid=0.9x1.25 maxpft=79 mask=gx1v7 use_cn=.true. use_nitrif_denitrif=.true. use_vertsoilc=.true. use_crop=.true. irrigate=.true. glc_nec=10
	</init_interp_attributes>

ctsm/bld/namelist_files/namelist_defaults_ctsm.xml

Lines 615 to 620 in ce3bab4

	<finidat hgrid="0.9x1.25" maxpft="79" mask="gx1v7" use_cn=".true." use_cndv=".false." use_fates=".false."
	ic_ymd="18500101" use_nitrif_denitrif=".true." use_vertsoilc=".true." sim_year="1850"
	ic_tod="0" glc_nec="10" use_crop=".true." irrigate=".true." use_init_interp=".true."
	lnd_tuning_mode="clm4_5_cam6.0"
	>lnd/clm2/initdata_map/clmi.B1850.0161-01-01.0.9x1.25_gx1v7_simyr1850_c190111.nc
	</finidat>

I'm going to locally edit the namelist file and see if this clears that subset of DIFF errors.

UPDATE: running with the local edit to irrigation in the namelist file result in the same DIFF errors. Did result in a new, unexpected RUN fail though. Suggests to me that irrigation difference is fine.

[glemieux]

Still waiting on the irrigate=.false. test check to run. I went and looked at the NLCOMP failures to see what was being reported in TestStatus.log. @ekluzek I'm seeing the following NLCOMP difference in the above list of DIFF fails that also have NLCOMP differences:

    Differences in namelist 'ndepdyn_nml':
    BASE: ndep_taxmode = 'cycle'
    COMP: ndep_taxmode = 'extend'
    BASE: stream_year_last_ndep = 2005
    COMP: stream_year_last_ndep = 2015

I'm not familiar with where the ndepdyn_nml contents get set. Do you think this update is acceptable?

UPDATE: I see it being initialized with extend via ndep_init in ndepStreamMod.F90 and some settings in user_default_ctsm.xml, but they match the ctsm1.0.dev093, so I'm guessing this is happening somewhere else that I'm not seeing.

I should note that the following only have the difference in ndep_taxmode and not a difference in stream_year_last_ndep:

ERS_Ly5_P72x1.f10_f10_musgs.IHistClm45BgcCrop.cheyenne_intel.clm-cropMonthOutput
SMS_D_Ly6_Mmpi-serial.1x1_smallvilleIA.IHistClm45BgcCropQianGs.cheyenne_intel.clm-cropMonthOutput

[glemieux]

Closing issue as the update has been complete via rebase.